Water level detection methods and devices, electronic equipment, and storage media

Abstract

This invention discloses a water level detection method, apparatus, electronic device, and storage medium, relating to the field of water level detection. The method includes: first, receiving an image captured by a target measuring tool, wherein the target measuring tool is set in the geographical environment of the water level to be measured; second, inputting the image into a specified neural network model, outputting character recognition results; third, evaluating the confidence level of the character recognition results; and fourth, extracting the least significant digit from the character recognition results if the confidence level is lower than or equal to a preset confidence threshold; and finally, outputting the water level value at the current moment based on the least significant digit. This invention solves the technical problem in related technologies where water level sensors cannot effectively detect water levels and are prone to false alarms.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of water level detection, in particular to a water level detection method and device, electronic equipment and storage medium. BACKGROUND

[0002] At present, many places are easily affected by extreme weather such as rainstorm and typhoon, and many areas have the risk of flood. In order to avoid losses, the water level of each area needs to be monitored and warned in time.

[0003] In the related art, when monitoring the water level of each area, a water level sensor is usually used for water level detection. However, the water level sensor is greatly affected by water quality during use, especially in the flood with a large amount of silt, which is prone to false positives and cannot accurately detect the water level. In addition, water level detection by the water level sensor also has the disadvantages of high cost and difficulty in maintenance.

[0004] At present, there is no effective solution to the above problems. SUMMARY

[0005] The embodiments of the present application provide a water level detection method and device, electronic equipment and storage medium, to at least solve the technical problem in the related art that the water level sensor cannot effectively detect the water level and is prone to false positives.

[0006] According to an aspect of an embodiment of the present application, a water level detection method is provided, comprising: receiving an image obtained by photographing a target measurement tool, wherein the target measurement tool is arranged in a geographical environment to be measured; inputting the image into a specified neural network model to output a character recognition result; evaluating the confidence of the character recognition result, and extracting the lowest value in the character recognition result if the confidence is lower than or equal to a preset confidence threshold; and outputting a water level value at the current time based on the lowest value.

[0007] Optionally, the step of inputting the image into a specified neural network model to output a character recognition result comprises: based on a water level preset bit, framing an area in the image associated with the target measurement tool, and taking a screenshot of the framed area to obtain an initial image of interest, wherein the water level preset bit is a fixed-point monitoring position of the target measurement tool in the geographical environment; the initial image of interest is rotated counterclockwise, and the rotated image is input into the specified neural network model to output a current reading set of the target measurement tool, thereby obtaining the character recognition result, wherein the current reading set at least includes the lowest value.

[0008] Optionally, the specified neural network model has been pre-trained, and the training of the specified neural network model comprises: obtaining a plurality of groups of historical photographed images, wherein each group of the historical photographed images comprises at least two images obtained by continuously photographing the target measuring tool in a historical time period; inputting the plurality of groups of historical photographed images into a convolutional neural network to extract image initial features; inputting the image initial features into a recurrent neural network to extract image sequence features; inputting the image sequence features into a preset loss function of connection time domain classification to train a convolutional recurrent neural network model; determining that the convolutional recurrent neural network model is trained when a similarity between a character recognition result output by the convolutional recurrent neural network model and an actual character value of the historical photographed images exceeds a preset similarity threshold; and taking the trained convolutional recurrent neural network model as the specified neural network model.

[0009] Optionally, the step of evaluating the confidence of the character recognition result comprises: extracting a minimum value and a maximum value in the character recognition result, and constructing continuous strings based on the minimum value and the maximum value, wherein a numerical interval between the continuous strings is referenced to a scale of the target measuring tool; calculating an edit distance value between the continuous strings and a current read collection in the character recognition result by using a preset edit distance formula, wherein the edit distance value is used to represent a modification times required for changing the current read collection to the continuous strings; and evaluating the confidence of the character recognition result based on the edit distance value.

[0010] Optionally, the step of outputting the water level value at the current time based on the lowest significant digit value comprises: obtaining a historical water level read collection, wherein the historical water level read collection comprises a plurality of historical water level values measured in a historical process; updating the historical water level read collection based on the lowest significant digit value to obtain a water level read collection of a current time period, wherein the current time period comprises the current time and a plurality of historical times before the current time; counting a number of occurrences of each water level value in the water level read collection of the current time period, and sorting all the numbers of occurrences to obtain a sorting result; extracting a maximum number of occurrences in the sorting result, and comparing the maximum number of occurrences with a preset numerical threshold to obtain a comparison result; in a case where the comparison result indicates that the maximum number of occurrences is greater than the preset numerical threshold, taking a water level value indicated by the maximum number of occurrences as the water level value at the current time; and in a case where the comparison result indicates that the maximum number of occurrences is less than or equal to the preset numerical threshold, taking a historical water level value closest to the current time in the historical water level read collection and recognized from a photographed image as the water level value at the current time.

[0011] Optionally, the step of updating the set of historical water level readings based on the least significant value to obtain a set of water level readings of the current time period comprises: determining whether all of the set of historical water level readings are null values, wherein the null value refers to an empty value assigned to a photographing time point after the photographing image fails to recognize a water level value; in the case that all of the set of historical water level readings are null values, substituting the least significant value into the set of historical water level readings to obtain the set of water level readings of the current time period; in the case that not all of the set of historical water level readings are null values, determining whether the least significant value is within a preset variable range of water level readings; in the case that the least significant value is within the preset variable range of water level readings, substituting the least significant value into the set of historical water level readings to obtain the set of water level readings of the current time period; in the case that the least significant value is not within the preset variable range of water level readings, substituting an empty value into the set of historical water level readings to obtain the set of water level readings of the current time period.

[0012] Optionally, before comparing the maximum occurrence number with the preset numerical threshold to obtain a comparison result, the method further comprises: obtaining a total number of all historical water level values in the set of historical water level readings; and determining the preset numerical threshold based on the total number of all historical water level values and the to-be-compared parameter.

[0013] According to another aspect of the embodiment of the present application, a water level detection device is also provided, comprising: a receiving unit configured to receive an image obtained by photographing a target measuring tool, wherein the target measuring tool is arranged in a geographical environment to be measured; an input unit configured to input the image into a specified neural network model to output a character recognition result; an evaluation unit configured to evaluate a confidence of the character recognition result, and extract a least significant value in the character recognition result in the case that the confidence is lower than or equal to a preset confidence threshold; and an output unit configured to output a water level value at a current time point based on the least significant value.

[0014] Optionally, the input unit comprises: a first screenshot module configured to frame an area associated with the target measuring tool in the image based on a water level preset position, and take a screenshot of the framed area to obtain an initial image of interest, wherein the water level preset position is a position at which the target measuring tool is arranged for monitoring in the geographical environment; and a first input module configured to rotate the initial image of interest counterclockwise, input the rotated image into the specified neural network model, and output a current reading set of the target measuring tool to obtain the character recognition result, wherein the current reading set at least comprises the least significant value.

[0015] Optionally, the water level detection apparatus further comprises: a first acquisition module, configured to acquire a plurality of groups of historical photographed images, wherein each group of the historical photographed images comprises at least two images obtained by continuously photographing the target measuring tool in a historical time period; a first extraction module, configured to input the plurality of groups of historical photographed images into a convolutional neural network to extract image initial features; a second extraction module, configured to input the image initial features into a recurrent neural network to extract image sequence features; a first training module, configured to input the image sequence features into a preset loss function of connection time domain classification to train a convolutional recurrent neural network model; a first determination module, configured to determine that the convolutional recurrent neural network model is trained when a similarity between a character recognition result output by the convolutional recurrent neural network model and an actual character value of the historical photographed images exceeds a preset similarity threshold; and a first serving module, configured to serve the trained convolutional recurrent neural network model as the specified neural network model.

[0016] Optionally, the evaluation unit comprises: a first construction module, configured to extract a minimum value and a maximum value in the character recognition result, and construct continuous character strings based on the minimum value and the maximum value, wherein a numerical interval between the continuous character strings is referenced to a scale of the target measuring tool; a first calculation module, configured to calculate an edit distance value between the continuous character strings and a current read collection in the character recognition result by using a preset edit distance formula, wherein the edit distance value is used to represent a modification times required for changing the current read collection to the continuous character strings; and a first evaluation module, configured to evaluate a confidence of the character recognition result based on the edit distance value.

[0017] Optionally, the output unit comprises: a second acquisition module, configured to acquire a historical water level reading set, wherein the historical water level reading set comprises a plurality of historical water level values measured in a historical process; a first update module, configured to update the historical water level reading set based on the least significant value to obtain a water level reading set of a current time period, wherein the current time period comprises the current moment and a plurality of historical moments before the current moment; a first statistical module, configured to count the number of occurrences of each water level value in the water level reading set of the current time period, and sort all the numbers of occurrences to obtain a sorting result; a first comparison module, configured to extract the maximum number of occurrences in the sorting result, and compare the maximum number of occurrences with a preset numerical threshold to obtain a comparison result; a second as module, configured to, in a case where the comparison result indicates that the maximum number of occurrences is greater than the preset numerical threshold, take the water level value indicated by the maximum number of occurrences as the water level value of the current moment; and a third as module, configured to, in a case where the comparison result indicates that the maximum number of occurrences is less than or equal to the preset numerical threshold, take the historical water level value in the historical water level reading set that is closest to the current moment and is recognized by the photographed image as the water level value of the current moment.

[0018] Optionally, the first update module comprises: a first judgment submodule, configured to judge whether all the historical water level reading set is empty value, wherein the empty value refers to an empty value assigned to the photographed moment after the water level value is not recognized by the photographed image; a first substitution submodule, configured to, in a case where all the historical water level reading set is empty value, substitute the least significant value into the historical water level reading set to obtain the water level reading set of the current time period; a second judgment submodule, configured to, in a case where all the historical water level reading set is not empty value, judge whether the least significant value is within a preset water level reading variable range; a second substitution submodule, configured to, in a case where the least significant value is within the preset water level reading variable range, substitute the least significant value into the historical water level reading set to obtain the water level reading set of the current time period; and a third substitution submodule, configured to, in a case where the least significant value is not within the preset water level reading variable range, substitute an empty value into the historical water level reading set to obtain the water level reading set of the current time period.

[0019] Optionally, the first update module further comprises: a first acquisition submodule, configured to acquire the total number of all historical water level values in the historical water level reading set; and a first determination submodule, configured to determine the preset numerical threshold based on the total number of all historical water level values and a to-be-compared parameter.

[0020] According to another aspect of the present invention, a computer-readable storage medium is also provided, the computer-readable storage medium including a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to perform any of the above-described water level detection methods.

[0021] According to another aspect of the present invention, an electronic device is also provided, including one or more processors and a memory, the memory being used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement any of the above-described water level detection methods.

[0022] In this disclosure, the following steps are adopted: first, an image obtained by taking a picture of the target measuring tool is received, wherein the target measuring tool is set in the geographical environment of the water level to be measured; then, the image is input into a specified neural network model, and the character recognition result is output; then, the confidence of the character recognition result is evaluated; and if the confidence is lower than or equal to a preset confidence threshold, the least significant bit value in the character recognition result is extracted; finally, the water level value at the current moment is output based on the least significant bit value.

[0023] In this disclosure, water level is measured using a preset measuring tool, and an image is captured by a camera. The water level image is then recognized based on a neural network model and character recognition technology. Based on the confidence level of the recognition result, the water level detection result is obtained accurately and effectively, improving the water level recognition rate and reducing errors. This solves the technical problem in related technologies where water level sensors cannot effectively detect water levels and are prone to false alarms. Attached Figure Description

[0024] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings:

[0025] Figure 1 This is a flowchart of an optional water level detection method according to an embodiment of the present invention;

[0026] Figure 2 This is a flowchart of an optional water level detection method according to an embodiment of the present invention;

[0027] Figure 3 This is an optional text recognition model training flowchart according to an embodiment of the present invention;

[0028] Figure 4 This is an optional confidence assessment flowchart according to an embodiment of the present invention;

[0029] Figure 5This is an optional delay processing flowchart according to an embodiment of the present invention;

[0030] Figure 6 This is a schematic diagram of an optional water level detection device according to an embodiment of the present invention;

[0031] Figure 7 This is a hardware structure block diagram of an electronic device (or mobile device) for a water level detection method according to an embodiment of the present invention. Detailed Implementation

[0032] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0033] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0034] It should be noted that the water level detection method and apparatus in this disclosure can be used in the field of water level detection to detect the current water level value, or in any field other than water level detection to detect the water level value. The application field of the water level detection method and apparatus in this disclosure is not limited.

[0035] It should be noted that all information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for display, data used for analysis, etc.) involved in this disclosure are information and data authorized by the user or fully authorized by all parties. For example, this system has an interface with relevant users or organizations. Before obtaining relevant information, it is necessary to send an acquisition request to the aforementioned user or organization through the interface, and obtain the relevant information after receiving consent information from the aforementioned user or organization.

[0036] This invention can be applied to various water level detection equipment / devices / products, especially water level detection systems that are configured / associated with image capturing equipment. It measures the water level based on a preset measuring tool, identifies the water level image based on a neural network model and character recognition technology, and evaluates the confidence level of the identification results, thereby accurately and effectively detecting the water level, improving the water level recognition rate, reducing errors, and also reducing costs.

[0037] The present invention will now be described in detail with reference to various embodiments.

[0038] Example 1

[0039] According to an embodiment of the present invention, a method for water level detection is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0040] Figure 1 This is a flowchart of an optional water level detection method according to an embodiment of the present invention, such as... Figure 1 As shown, the method includes the following steps:

[0041] Step S102: Receive the image obtained by taking a picture of the target measuring tool, wherein the target measuring tool is set in the geographical environment of the water level to be measured;

[0042] Step S104: Input the image into the specified neural network model and output the character recognition result;

[0043] Step S106: Evaluate the confidence level of the character recognition result, and extract the least significant digit from the character recognition result if the confidence level is lower than or equal to a preset confidence threshold.

[0044] Step S108: Output the water level value at the current moment based on the least significant bit value.

[0045] Through the above steps, the system first receives images captured by a target measuring tool, which is set in the geographical environment where the water level to be measured. Then, the images are input into a specified neural network model, which outputs character recognition results. The confidence level of the character recognition results is then evaluated. If the confidence level is lower than or equal to a preset confidence threshold, the least significant digit of the character recognition result is extracted. Finally, the water level value at the current moment is output based on the least significant digit. In this embodiment, water level is measured using a preset measuring tool, and images of the measuring tool are captured by an imaging device. The water level image is then recognized based on a neural network model and character recognition technology. Based on the evaluated confidence level of the recognition results, the water level detection result is obtained accurately and effectively, improving the water level recognition rate and reducing errors. This solves the technical problem in related technologies where water level sensors cannot effectively detect water levels and are prone to false alarms.

[0046] The following section will explain in detail each of the above implementation steps.

[0047] Step S102: Receive the image obtained by taking a picture of the target measuring tool, wherein the target measuring tool is set in the geographical environment of the water level to be measured.

[0048] It should be noted that the implementing entity of this invention is a server. The server communicates with the camera via a network, and the camera uploads the captured data to the server for analysis. The target measuring tool is a continuous digital water gauge, which is pre-positioned at the water level to be measured. The camera is then used to capture images of the data on the digital water gauge to obtain images containing water level data. The geographical environment of the water level to be measured includes, but is not limited to: the deepest part of the river, the shallowest part of the river, the area where river sediment accumulates, and the lowest point in the park.

[0049] Optionally, after obtaining the water level image, due to the special geographical location of the shooting, the captured image may be unclear. Therefore, the image is first preprocessed by performing median filtering and image equalization to remove local noise and enhance contrast, so as to obtain a preprocessed image.

[0050] In this embodiment of the invention, before character recognition, a specified neural network model needs to be pre-trained. Training the specified neural network model includes: acquiring multiple sets of historically captured images, wherein each set of historically captured images contains at least two images obtained by continuously capturing the target measurement tool over a historical time period; inputting the multiple sets of historically captured images into a convolutional neural network to extract initial image features; inputting the initial image features into a recurrent neural network to extract image sequence features; inputting the image sequence features into a preset loss function for connecting temporal classification to train the convolutional recurrent neural network model; determining that the training of the convolutional recurrent neural network model is complete when the similarity between the character recognition result output by the convolutional recurrent neural network model and the actual character values ​​of the historically captured images exceeds a preset similarity threshold; and using the trained convolutional recurrent neural network model as the specified neural network model.

[0051] It should be noted that the specified neural network model is used to identify images captured by the camera to obtain water level data. When training the neural network model, multiple historical images need to be acquired in advance. From these images, the water level gauge portion is extracted according to a preset position and rotated by a specified angle (e.g., 270 degrees clockwise or 90 degrees counterclockwise) before being input into the pre-trained convolutional recurrent neural network (CNN) for inference to obtain the water level gauge reading. It is important to note that the pre-trained CNN model requires a large number of water level gauge images to be input into the CNN for initial feature extraction. The extracted features are then input into the CNN to extract sequence features, which are then input into the connection layer. Finally, the sequence features are input into a preset loss function for temporal classification to train the network. After training with a large number of images until the network converges, the pre-trained model is obtained. Alternatively, the training of the CNN model is considered complete when the similarity between the character recognition result output by the CNN model and the actual character values ​​in the historical images exceeds a preset similarity threshold.

[0052] Step S104: Input the image into the specified neural network model and output the character recognition result.

[0053] In this embodiment of the invention, step S104 includes: selecting a region in the image associated with the target measuring tool based on a preset water level position, and taking a screenshot of the selected region to obtain an initial image of interest, wherein the preset water level position is used to set the fixed monitoring position of the target measuring tool in the geographical environment; rotating the initial image of interest counterclockwise, and inputting the rotated image into a specified neural network model to output the current reading set of the target measuring tool to obtain a character recognition result, wherein the current reading set includes at least the least significant bit value.

[0054] Step S106: Evaluate the confidence level of the character recognition result, and extract the least significant digit from the character recognition result if the confidence level is lower than or equal to a preset confidence threshold.

[0055] It should be noted that due to factors such as lighting and occlusion, the recognized readings may be repeated or missing. Therefore, it is necessary to evaluate the confidence level of the obtained character recognition results. If the confidence level is less than or equal to the preset confidence threshold, the recognition result is considered reliable, and the least significant bit of the current character is output to obtain the value of the least significant bit; otherwise, the recognition is considered to be interfered with, and the null value "None" is output.

[0056] In this embodiment of the invention, step S106 includes: extracting the minimum and maximum values ​​from the character recognition results, and constructing a continuous string based on the minimum and maximum values, wherein the numerical interval between the continuous strings is referenced to the scale of the target measuring tool; calculating the edit distance value between the continuous string and the current reading set in the character recognition results using a preset edit distance formula, wherein the edit distance value is used to characterize the number of modifications required to change the current reading set into a continuous string; and evaluating the confidence level of the character recognition results based on the edit distance value.

[0057] Optionally, when performing confidence assessment, a continuous string is first constructed based on the highest and lowest bits of the character extraction result. For example, if the character extraction result is denoted as P and the reading is "9876542", then the constructed continuous string S is "98765432". The preset edit distance between the two is calculated as follows:

[0058]

[0059] Among them, lwv P，S The edit distance between the character extraction result P and the constructed continuous string S is represented by i, where i represents the i-th character in the character extraction result P and j represents the j-th character in the constructed continuous string S.

[0060] Edit distance calculates the number of modifications required to transform string P into string S. The edit distance value is determined by the number of modifications. If the extracted character result P is "9876542" and the constructed continuous string S is "98765432", then the edit distance between the two is lev. P，S =1, compare it with the confidence threshold. If the confidence threshold is set to T=1, then this reading is valid, and the lowest value is output as the minimum water level. When the character extraction result P is "987652", and the constructed continuous string S is "98765432", then the edit distance between the two is lev. P，S =2, compare it with the confidence threshold T=1. Since the edit distance is greater than the confidence threshold, the reading is considered to be disturbed, and the null value "None" is output.

[0061] Step S108: Output the water level value at the current moment based on the least significant bit value.

[0062] In this embodiment of the invention, the step of updating the historical water level reading set based on the least significant bit value to obtain the water level reading set for the current time period includes: determining whether all values ​​in the historical water level reading set are empty, wherein an empty value refers to an empty value assigned to the shooting time after the water level value was not recognized in the captured image; if all values ​​in the historical water level reading set are empty, substituting the least significant bit value into the historical water level reading set to obtain the water level reading set for the current time period; if not all values ​​in the historical water level reading set are empty, determining whether the least significant bit value is within a preset variable range of water level readings; if the least significant bit value is within the preset variable range of water level readings, substituting the least significant bit value into the historical water level reading set to obtain the water level reading set for the current time period; if the least significant bit value is not within the preset variable range of water level readings, substituting an empty value into the historical water level reading set to obtain the water level reading set for the current time period.

[0063] It should be noted that before outputting the current water level value, it is necessary to collect the water level readings over a historical period. Specifically, the readings of the previous n frames are obtained to get the historical water level reading set L. n ={L i-n L i-n+ 1......L i-1}

[0064] When L n When not all values ​​are "None", since water level changes are generally continuous, it is necessary to limit the variable range of water level readings. If the base value of the variable range is set to c, then the variable range is:

[0065] R = [min(L) n )-c,max(L n )+c];

[0066] For example, the historical water level reading obtained is L n ={1,1,2,2,2,}, c is set to 1, then R = [0,3]. Therefore, water level readings between 0 and 3 are considered normal. Add L to L n If the reading is not correct, then it is considered an abnormal reading, and "None" will be added to L. n middle.

[0067] If L n When all values ​​in L are "None", there is no need to limit the reading range; simply add the reading L directly to L. n In the middle (corresponding to the above, substituting the least significant value L into the historical water level reading set L),n ).

[0068] In this embodiment of the invention, step S108 includes: acquiring a historical water level reading set, wherein the historical water level reading set contains multiple historical water level values ​​measured during the historical process; updating the historical water level reading set based on the least significant value to obtain a water level reading set for the current time period, wherein the current time period includes: the current moment and multiple historical moments before the current moment; counting the occurrence frequency of each water level value in the water level reading set for the current time period, and sorting all occurrence frequencies to obtain a sorting result; extracting the maximum occurrence frequency in the sorting result, and comparing the maximum occurrence frequency with a preset numerical threshold to obtain a comparison result; if the comparison result indicates that the maximum occurrence frequency is greater than the preset numerical threshold, taking the water level value indicated by the maximum occurrence frequency as the water level value at the current moment; if the comparison result indicates that the maximum occurrence frequency is less than or equal to the preset numerical threshold, taking the historical water level value identified by the captured image at the moment closest to the current moment in the historical water level reading set as the water level value at the current moment.

[0069] It should be noted that before comparing the maximum occurrence frequency with the preset numerical threshold to obtain the comparison result, the process also includes: obtaining the total number of all historical water level values ​​in the historical water level reading set; and determining the preset numerical threshold based on the total number of all historical water level values ​​and the comparison parameter. Let n be the total number of all historical water level values, which can be 2 or 3 as the comparison parameter. For example, the preset numerical threshold can be determined to be half or one-third of n.

[0070] Optionally, before outputting the final water level value, it is also necessary to calculate multiple water level values ​​L within the historical time period. n The most frequent reading L b The number of times b (the maximum number of occurrences in the sorted results) is used. If b is greater than half of n (a preset numerical threshold), then the reading is considered valid, and the final water level value L is output. b Otherwise, the current reading is considered invalid, and the previous valid reading value is output.

[0071] Optionally, in this embodiment, when outputting the final water level value, a historical reading list is maintained to determine whether the current reading result is continuous. If it is continuous, the reading with the maximum number of occurrences within the specified time window is output as the final reading; otherwise, the last valid final reading is output.

[0072] Through the above embodiments, based on character recognition technology, confidence assessment technology, and delay processing technology, the captured data is processed to obtain the final water level value, thereby accurately and effectively detecting the water level, improving the water level recognition rate, reducing errors, and also reducing costs.

[0073] The invention will now be described in conjunction with another alternative embodiment.

[0074] Example 2

[0075] This embodiment provides a water level detection system, and the various implementation modules included in the water level detection system correspond to the various implementation steps in Embodiment 1.

[0076] The water level detection system uses character recognition technology to identify and process water level values. It includes four core modules: information acquisition module, information recognition module, confidence module, and delay processing module. Its cost is much lower than that of traditional water level sensor technology. Moreover, water level sensors are greatly affected by water quality in actual use, especially floods with a lot of silt, which are prone to false alarms. The embodiment of this invention has a high water level recognition rate, smaller error, and is easier to maintain, making it more suitable for widespread promotion and use.

[0077] Figure 2 This is a flowchart of an optional water level detection method according to an embodiment of the present invention, such as... Figure 2 As shown, the water level detection method includes the following steps:

[0078] Step 1: Begin;

[0079] Step 2: Take a picture of the water level gauge;

[0080] In this embodiment of the invention, when acquiring images, a digital water gauge is placed at a designated location for measuring water level, and a webcam is used to continuously capture images of it and transmit the acquired images I to the server.

[0081] Step 3: Input the image into the text recognition model and obtain the recognition result P;

[0082] When performing image information recognition Figure 3 This is an optional text recognition model training flowchart according to an embodiment of the present invention, such as... Figure 3 As shown, the training process for this text recognition model is as follows:

[0083] Step 301: Begin;

[0084] Step 302: Obtain water level image;

[0085] Step 303: Input into the convolutional recurrent neural network;

[0086] Step 304: Input to the connection-temporal classification loss function;

[0087] Step 305: End.

[0088] Through the above steps, the water level gauge portion is extracted from the acquired image I according to the preset position, and then rotated 90 degrees counterclockwise before being input into the pre-trained convolutional recurrent neural network for inference to obtain the water level gauge reading P. The pre-trained convolutional recurrent neural network model needs to input a large number of water level gauge images into the convolutional neural network to initially extract features, and then input the extracted features into the recurrent neural network to extract sequence features F. Then, the extracted sequence features F are input into the connection layer, and finally input into the connection-time classification (CTC) loss function for network training. After training with a large number of images until the network converges, the pre-trained text recognition model can be obtained.

[0089] The character recognition model is trained in advance based on the above steps. When the information acquisition module sends the acquired water level image to the server, the information recognition module calls the image for recognition. It takes a screenshot based on the preset water level position to obtain the initial image, rotates it counterclockwise, and inputs it into the character recognition model to obtain the character recognition result P.

[0090] Step 4: Input P into the confidence module to evaluate the confidence level and take the least significant digit as the current water level L;

[0091] Due to factors such as lighting and occlusion, the recognized readings may be repeated or missing. Therefore, it is necessary to evaluate the confidence level of the obtained character recognition results. If the confidence level is less than or equal to the preset confidence threshold, the recognition result is considered reliable, and the least significant bit of the current character is output; otherwise, the recognition is considered to have been interfered with, and "None" is output. Figure 4 This is an optional confidence assessment flowchart according to an embodiment of the present invention, such as... Figure 4 As shown, the confidence assessment process includes:

[0092] Step 401: Begin;

[0093] Step 402: Input the text recognition result P;

[0094] Step 403: Construct string S;

[0095] Step 404: Calculate the edit distance between P and S;

[0096] Step 405: Determine whether it is less than or equal to the threshold M. If yes, proceed to step 406; otherwise, proceed to step 407.

[0097] Step 406: Output the least significant bit L of P;

[0098] Step 407: Output "None";

[0099] Step 408: End.

[0100] When conducting confidence assessment, a continuous string is first constructed based on the highest and lowest bits of the character extraction result. For example, if the character extraction result is denoted as P and the reading is "9876542", then the constructed continuous string S is "98765432". The formula for calculating the edit distance between the two is as follows:

[0101]

[0102] Among them, lev P，s The edit distance between the character extraction result P and the constructed continuous string S is represented by i, where i represents the i-th character in the character extraction result P and j represents the j-th character in the constructed continuous string S.

[0103] Edit distance calculates the number of modifications required to transform string P into string S. If the extracted string P is "9876542" and the constructed continuous string S is "98765432", then the edit distance between the two is lev. P，S =1, compare it with the confidence threshold. If the confidence threshold is set to T=1, then this reading is valid, and the lowest value is output as the minimum water level. When the character extraction result P is "987652", and the constructed continuous string S is "98765432", then the edit distance between the two is lev. P，S =2, compare it with the confidence threshold T=1. Since the edit distance is greater than the confidence threshold, the reading is considered to be disturbed, and "None" is output.

[0104] Step 5: Input L into the delay processing module and output the final water level result F;

[0105] Figure 5 This is an optional delay processing flowchart according to an embodiment of the present invention, such as... Figure 5 As shown, the delay processing flow includes:

[0106] Step 501: Begin;

[0107] Step 502: Input the current water level reading L and the historical water level reading result L n ;

[0108] Step 503: Determine if it is within the allowed range. If yes, proceed to step 504; otherwise, proceed to step 505.

[0109] Step 504: Add L to L n middle;

[0110] Step 505: Add None to L n middle;

[0111] Step 506: Calculate L nThe reading L that appears most frequently b The number of times b;

[0112] Step 507: Determine if b is greater than N / 2. If yes, proceed to step 508; otherwise, proceed to step 509.

[0113] Step 508: Output the final water level value;

[0114] Step 509: Output the previous valid water level value;

[0115] Step 510: End.

[0116] Based on the above steps, before outputting the current water level value, it is necessary to statistically analyze the water level readings over a historical time period. Specifically, this involves obtaining the readings L from the previous n frames of the current frame i. n ={L i-n L i-n+1 ......L i-1}

[0117] When L n When not all values ​​are "None", since water level changes are generally continuous, it is necessary to limit the variable range of water level readings. If the base value of the variable range is set to c, then the variable range is:

[0118] R = [min(L) n )-c,max(L n )+c];

[0119] For example, the obtained historical water level reading is L n ={1,1,2,2,2,}, c is set to 1, then R = [0,3]. Therefore, water level readings between 0 and 3 are considered normal. Add L to L n If the reading is not correct, then it is considered an abnormal reading, and "None" will be added to L. n middle.

[0120] If L n When all values ​​in L are "None", there is no need to limit the reading range; simply add the reading L directly to L. n middle.

[0121] Finally, before outputting the final water level value, it is also necessary to calculate multiple water level values ​​L within the historical time period. n The most frequent reading L b The number of times b is taken. If b is greater than half of n, then the reading is considered valid, and the final water level value is output as L. b Otherwise, the current reading is considered invalid, and the previous valid reading value is output.

[0122] Step Six: End.

[0123] Through the above embodiments, based on character recognition technology, confidence assessment technology, and delay processing technology, the captured data is processed to obtain the final water level value, thereby accurately and effectively detecting the water level, improving the water level recognition rate, reducing errors, and also reducing costs.

[0124] The invention will now be described in conjunction with another alternative embodiment.

[0125] Example Three

[0126] This embodiment provides a water level detection device, and the various implementation units included in the water level detection device correspond to the various implementation steps in Embodiment 1.

[0127] Figure 6 This is a schematic diagram of an optional water level detection device according to an embodiment of the present invention, such as... Figure 6 As shown, the water level detection device includes: a receiving unit 62, an input unit 64, an evaluation unit 66, and an output unit 68, wherein,

[0128] The receiving unit 62 is used to receive images captured by the target measuring tool, wherein the target measuring tool is set in the geographical environment of the water level to be measured;

[0129] Input unit 64 is used to input images into a specified neural network model and output character recognition results;

[0130] Evaluation unit 66 is used to evaluate the confidence level of the character recognition result, and extract the least significant bit value in the character recognition result if the confidence level is lower than or equal to a preset confidence threshold.

[0131] Output unit 68 is used to output the water level value at the current moment based on the least significant bit value.

[0132] The aforementioned water level detection device receives images captured by a target measuring tool via a receiving unit 62, wherein the target measuring tool is set in the geographical environment of the water level to be measured; the image is input to a specified neural network model via an input unit 64, and the character recognition result is output; the confidence level of the character recognition result is evaluated via an evaluation unit 66, and if the confidence level is lower than or equal to a preset confidence threshold, the least significant bit value in the character recognition result is extracted; the water level value at the current moment is output via an output unit 68 based on the least significant bit value.

[0133] In this embodiment, the water level is measured using a preset measuring tool, and an image is captured by a camera. The water level image is then recognized based on a neural network model and character recognition technology. Based on the confidence level of the recognition result, the water level detection result is obtained accurately and effectively, improving the water level recognition rate and reducing errors. This solves the technical problem in related technologies where water level sensors cannot effectively detect water levels and are prone to false alarms.

[0134] Optionally, the input unit 64 includes: a first screenshot module, used to select a region of the image associated with the target measuring tool based on a preset water level position, and take a screenshot of the selected region to obtain an initial image of interest, wherein the preset water level position is used to set the fixed monitoring position of the target measuring tool in the geographical environment; and a first input module, used to rotate the initial image of interest counterclockwise, input the rotated image into a specified neural network model, output the current reading set of the target measuring tool, and obtain a character recognition result, wherein the current reading set includes at least the least significant bit value.

[0135] Optionally, the water level detection device further includes: a first acquisition module for acquiring multiple sets of historical images, wherein each set of historical images contains at least two images obtained by continuously capturing the target measuring tool over a historical time period; a first extraction module for inputting the multiple sets of historical images into a convolutional neural network to extract initial image features; a second extraction module for inputting the initial image features into a recurrent neural network to extract image sequence features; a first training module for inputting the image sequence features into a preset loss function for connecting temporal classification to train the convolutional recurrent neural network model; a first determination module for determining the trained convolutional recurrent neural network model when the similarity between the character recognition result output by the convolutional recurrent neural network model and the actual character values ​​of the historical images exceeds a preset similarity threshold; and a first use module for using the trained convolutional recurrent neural network model as a specified neural network model.

[0136] Optionally, the evaluation unit 66 includes: a first construction module, used to extract the minimum and maximum values ​​in the character recognition results, and construct a continuous string based on the minimum and maximum values, wherein the numerical interval between the continuous strings is referenced to the scale of the target measuring tool; a first calculation module, used to calculate the edit distance value between the continuous string and the current reading set in the character recognition results using a preset edit distance formula, wherein the edit distance value is used to characterize the number of modifications required to change the current reading set into a continuous string; and a first evaluation module, used to evaluate the confidence level of the character recognition results based on the edit distance value.

[0137] Optionally, the output unit 68 includes: a second acquisition module for acquiring a set of historical water level readings, wherein the set of historical water level readings contains multiple historical water level values ​​measured during the historical process; a first update module for updating the set of historical water level readings based on the least significant value to obtain a set of water level readings for the current time period, wherein the current time period includes: the current moment and multiple historical moments before the current moment; a first statistics module for counting the occurrence frequency of each water level value in the set of water level readings for the current time period and sorting all occurrence frequencies to obtain a sorting result; a first comparison module for extracting the maximum occurrence frequency in the sorting result and comparing the maximum occurrence frequency with a preset numerical threshold to obtain a comparison result; a second use module for using the water level value indicated by the maximum occurrence frequency as the water level value of the current moment when the comparison result indicates that the maximum occurrence frequency is greater than the preset numerical threshold; and a third use module for using the historical water level value identified by the captured image from the set of historical water level readings that is closest to the current moment as the water level value of the current moment when the comparison result indicates that the maximum occurrence frequency is less than or equal to the preset numerical threshold.

[0138] Optionally, the first update module includes: a first judgment submodule, used to determine whether all values ​​in the historical water level reading set are empty, wherein an empty value refers to an empty value assigned to the shooting time after the water level value was not recognized in the captured image; a first substitution submodule, used to substitute the lowest value into the historical water level reading set to obtain the water level reading set for the current time period when all values ​​in the historical water level reading set are empty; a second judgment submodule, used to determine whether the lowest value is within a preset variable range of water level readings when not all values ​​in the historical water level reading set are empty; a second substitution submodule, used to substitute the lowest value into the historical water level reading set to obtain the water level reading set for the current time period when the lowest value is within the preset variable range of water level readings; and a third substitution submodule, used to substitute an empty value into the historical water level reading set to obtain the water level reading set for the current time period when the lowest value is not within the preset variable range of water level readings.

[0139] Optionally, the first update module further includes: a first acquisition submodule, used to acquire the total number of all historical water level values ​​in the historical water level reading set; and a first determination submodule, used to determine a preset numerical threshold based on the total number of all historical water level values ​​and the parameter to be compared.

[0140] The water level detection device described above may also include a processor and a memory. The receiving unit 62, input unit 64, evaluation unit 66, output unit 68, etc., are all stored in the memory as program units, and the processor executes the program units stored in the memory to realize the corresponding functions.

[0141] The processor described above contains a kernel, which retrieves the corresponding program unit from memory. One or more kernels can be configured, and the water level value is detected by adjusting the kernel parameters.

[0142] The aforementioned memory may include non-permanent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0143] According to another aspect of the present invention, a computer-readable storage medium is also provided, the computer-readable storage medium including a stored computer program, wherein, when the computer program is running, it controls the device where the computer-readable storage medium is located to execute any of the above-described water level detection methods.

[0144] According to another aspect of the present invention, an electronic device is also provided, including one or more processors and a memory, the memory being used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement any of the above-described water level detection methods.

[0145] This application also provides a computer program product, which, when executed on a data processing device, is suitable for executing a program that initializes the following method steps:

[0146] Figure 7 This is a hardware structure block diagram of an electronic device (or mobile device) for a water level detection method according to an embodiment of the present invention. Figure 7 As shown, the electronic device may include one or more processors 702 (shown as 702a, 702b, ..., 702n in the figure) 702 (processor 702 may include, but is not limited to, a microprocessor MCU or a programmable logic device FPGA, etc.) and a memory 704 for storing data. In addition, it may include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of the I / O interface), a network interface, a keyboard, a power supply, and / or a camera. Those skilled in the art will understand that... Figure 7 The structure shown is for illustrative purposes only and does not limit the structure of the electronic device described above. For example, the electronic device may also include components that are more... Figure 7 The more or fewer components shown, or having the same Figure 7 The different configurations shown.

[0147] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0148] In the above embodiments of the present invention, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0149] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0150] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0151] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0152] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0153] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A water level detecting method characterized by comprising: The method comprises: receiving an image taken by a target measuring tool, wherein the target measuring tool is arranged in a geographical environment to be measured; inputting the image into a specified neural network model to output a character recognition result, wherein the specified neural network model has been pre-trained, and the training of the specified neural network model comprises: obtaining a plurality of groups of historical photographed images, wherein each group of the historical photographed images contains at least two images taken by the target measuring tool in a historical time period; inputting the plurality of groups of historical photographed images into a convolutional neural network to extract image initial features; inputting the image initial features into a recurrent neural network to extract image sequence features; inputting the image sequence features into a preset loss function of connection time domain classification to train a convolutional recurrent neural network model; determining that the convolutional recurrent neural network model is trained when a similarity between a character recognition result output by the convolutional recurrent neural network model and an actual character value of the historical photographed images exceeds a preset similarity threshold; and taking the trained convolutional recurrent neural network model as the specified neural network model; evaluating a confidence of the character recognition result, and extracting a lowest value in the character recognition result when the confidence is lower than or equal to a preset confidence threshold; the step of evaluating the confidence of the character recognition result comprises: extracting a minimum value and a maximum value in the character recognition result, and constructing continuous strings based on the minimum value and the maximum value, wherein a numerical interval between the continuous strings is referenced to a scale of the target measuring tool; calculating an edit distance value between the continuous strings and a current reading set in the character recognition result by using a preset edit distance formula, wherein the edit distance value is used to represent a modification times required for changing the current reading set to the continuous strings; and evaluating the confidence of the character recognition result based on the edit distance value. outputting a water level value at the current moment based on the least significant digit value, including: obtaining a historical water level reading set, wherein the historical water level reading set contains a plurality of historical water level values measured in a historical process; updating the historical water level reading set based on the least significant digit value to obtain a water level reading set of a current time period, wherein the current time period contains the current moment and a plurality of historical moments before the current moment; counting the number of occurrences of each water level value in the water level reading set of the current time period, and sorting all the numbers of occurrences to obtain a sorting result; extracting the maximum number of occurrences in the sorting result, and comparing the maximum number of occurrences with a preset numerical threshold to obtain a comparison result; in a case where the comparison result indicates that the maximum number of occurrences is greater than the preset numerical threshold, taking the water level value indicated by the maximum number of occurrences as the water level value at the current moment; in a case where the comparison result indicates that the maximum number of occurrences is less than or equal to the preset numerical threshold, taking a historical water level value in the historical water level reading set that is closest to the current moment and is recognized from a photographed image as the water level value at the current moment.

2. The method of claim 1, wherein, The step of inputting the image into a specified neural network model and outputting a character recognition result includes: based on a water level preset bit, the region in the image associated with the target measurement tool is framed and the framed region is captured to obtain an initial image of interest, wherein the water level preset bit is to set the fixed-point monitoring position of the target measurement tool in the geographical environment; the initial image of interest is rotated counterclockwise, and the rotated image is input into the specified neural network model to output a current reading set of the target measurement tool, and the character recognition result is obtained, wherein the current reading set at least includes the least significant digit value.

3. The method of claim 1, wherein, The step of updating the historical water level reading set based on the least significant digit value to obtain a water level reading set of a current time period includes: determining whether all the historical water level reading set is empty, wherein the empty value refers to an empty value assigned to the shooting moment when the water level value is not recognized from the photographed image; in a case where all the historical water level reading set is empty, the least significant digit value is substituted into the historical water level reading set to obtain the water level reading set of the current time period; in a case where not all the historical water level reading set is empty, it is determined whether the least significant digit value is within a pre-set water level reading variable range; in a case where the least significant digit value is within the pre-set water level reading variable range, the least significant digit value is substituted into the historical water level reading set to obtain the water level reading set of the current time period; in a case where the least significant digit value is not within the pre-set water level reading variable range, an empty value is substituted into the historical water level reading set to obtain the water level reading set of the current time period.

4. The method of claim 1, wherein, Before comparing the maximum number of occurrences with the preset numerical threshold to obtain the comparison result, it further includes: obtaining the total number of all historical water level values in the historical water level reading set; Determine the preset numerical threshold value based on the total number of all historical water level values and the to-be-compared parameter.

5. A water level detecting device characterized by comprising: Comprise: A receiving unit configured to receive an image obtained by photographing a target measurement tool, wherein the target measurement tool is arranged in a geographical environment to be measured; An input unit configured to input the image into a designated neural network model to output a character recognition result; and an evaluation unit configured to evaluate a confidence level of the character recognition result, and extract a lowest numerical value in the character recognition result if the confidence level is lower than or equal to a preset confidence threshold value; The evaluation unit comprises: a first construction module configured to extract a minimum numerical value and a maximum numerical value in the character recognition result, and construct continuous strings based on the minimum numerical value and the maximum numerical value, wherein the numerical interval between the continuous strings is referenced to the scale of the target measurement tool; a first calculation module configured to calculate an edit distance value between the continuous strings and a current reading set in the character recognition result by using a preset edit distance formula, wherein the edit distance value is used to represent the number of modifications required to change the current reading set to the continuous strings; and a first evaluation module configured to evaluate the confidence level of the character recognition result based on the edit distance value; An output unit configured to output a water level value at a current time based on the lowest numerical value; The output unit comprises: a second acquisition module configured to acquire a historical water level reading set, wherein the historical water level reading set comprises a plurality of historical water level values measured in a historical process; a first update module configured to update the historical water level reading set based on the lowest numerical value to obtain a water level reading set of a current time period, wherein the current time period comprises the current time and a plurality of historical times before the current time; a first statistical module configured to count the number of occurrences of each water level value in the water level reading set of the current time period, sort all the numbers of occurrences to obtain a sorting result; a first comparison module configured to extract a maximum number of occurrences in the sorting result, and compare the maximum number of occurrences with a preset numerical threshold value to obtain a comparison result; a second as module configured to, if the comparison result indicates that the maximum number of occurrences is greater than the preset numerical threshold value, take the water level value indicated by the maximum number of occurrences as the water level value at the current time; and a third as module configured to, if the comparison result indicates that the maximum number of occurrences is less than or equal to the preset numerical threshold value, take a historical water level value in the historical water level reading set that is closest to the current time and is recognized from the photographed image as the water level value at the current time. The water level detection device further comprises: a first acquisition module, configured to acquire a plurality of groups of historical shooting images, wherein each group of the historical shooting images comprises at least two images obtained by continuously shooting the target measuring tool in a historical time period; a first extraction module, configured to input the plurality of groups of historical shooting images into a convolutional neural network to extract image initial features; a second extraction module, configured to input the image initial features into a recurrent neural network to extract image sequence features; a first training module, configured to input the image sequence features into a preset loss function of connection time domain classification to train a convolutional recurrent neural network model; a first determination module, configured to determine that the convolutional recurrent neural network model is trained when a similarity between a character recognition result output by the convolutional recurrent neural network model and an actual character value of the historical shooting image exceeds a preset similarity threshold; and a first serving module, configured to serve the trained convolutional recurrent neural network model as the specified neural network model.

6. A computer-readable storage medium, characterized in that, The computer readable storage medium comprises a stored computer program, wherein the computer readable storage medium controls a device in which the computer readable storage medium is located to perform the water level detection method of any one of claims 1 to 4 when the computer program is running.

7. An electronic device, comprising: The device comprises one or more processors and a memory, and the memory is configured to store one or more programs, wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the water level detection method of any one of claims 1 to 4.

Images